India operates 1,000+ central and state welfare schemes — income support, scholarships, pensions, housing, health insurance, disability benefits. The funds are allocated. The policies exist. The problem is discovery and access.
- No single interface exists for a citizen to ask "what am I eligible for?"
- Eligibility criteria are buried across hundreds of PDFs, portals, and departments in dense legalese
- Even when citizens find a scheme, they don't know which documents are missing or how to apply
The result: billions of rupees in welfare benefits go unclaimed every year — by exactly the people they were designed to reach.
Yojana Saathi collapses that entire journey into one application. Sign up, input your profile, and receive a personalized dashboard showing exactly what you qualify for — complete with missing document detection, OCR-assisted verification, and pre-filled application drafts.
|
Deterministic Eligibility Engine A pure rules-based engine matches citizen profiles against scheme criteria with auditable 0–1 confidence scores. No AI hallucinations for legal entitlements. |
Document Verification via OCR Upload Aadhaar, income certificates, or ration cards. OCR extracts parameters automatically — citizens verify before data enters their profile. |
AI-Drafted Applications One click generates a pre-filled application letter for any matched scheme — personalized with verified citizen data and scheme-specific requirements. |
Privacy-First Architecture Row-Level Security on every table. Zero PII in logs. Short-lived signed URLs for documents. The system runs fully correctly with no API key at all. |
Every eligibility decision flows through a deterministic, fully auditable pipeline — no probabilistic AI touches legal entitlements.
🔒 Core Design Principle: The LLM is never asked "is this person eligible?" — that judgment touches real legal entitlements and must be reproducible. Gemini is used only to polish benefit summaries and draft application letters, with timeout-based fallback to deterministic template text on any failure.
| Layer | Technologies |
|---|---|
| Backend API |     |
| Frontend |     |
| Database & Auth |    |
| AI & OCR |   |
| DevOps |    |
| Quality | -0A9EDC?style=flat-square&logo=pytest&logoColor=white){kind=icon}   |
- Deterministic 6-agent eligibility pipeline with auditable confidence scoring
- Secure citizen dashboard with ranked scheme matching
- Document vault with OCR verification and parameter extraction
- AI-polished application draft generation with graceful degradation
- 119 automated tests + GitHub Actions CI pipeline
- Multi-stage Docker deployment on Render (Infrastructure-as-Code)
- Comprehensive security hardening — CORS allowlist, rate limiting, input sanitization
- Multilingual support (Hindi + regional languages)
- Admin console UI for dynamic scheme and rule management
- Expanded scheme dataset beyond initial 20 for broader central & state coverage
- Analytics dashboard for benefit distribution and gap analysis
- DigiLocker API integration for automated document verification
- NGO and field-agent deployment toolkit
- Native mobile applications (Android / iOS)
- Government sandbox integrations for real-time verification
Prem Rajesh Sargara Founder · Backend & Architecture Agent pipeline · API orchestration · CI/CD · Deployment 
|
Dvij Joshi Co-Founder · Frontend & Product UI/UX · Auth · Dashboard · Document vault · Chat 
|
We're building public infrastructure in the open. Whether you write code, research policy, translate documents, or design interfaces — there's meaningful work here.
| 🐛 Find a starting point | Good First Issues |
| 💬 Join the conversation | GitHub Discussions |
| 📖 Read the playbook | Contributing Guide |
| 🛡️ Report a vulnerability | Security Policy |
| 📋 Understand the rules | Code of Conduct |
| 🏗️ Deep dive architecture | Project Wiki |
| Repository | Description | Status |
|---|---|---|
| Yojana-Saathi | Core platform — FastAPI backend + Next.js frontend |  |
| .github | Organization profile, branding & governance | Active |
